A method and system for simulating water pollution migration and transformation in a watershed based on an MT-DHM model

Abstract

The present application relates to the technical field of migration simulation, and more particularly to a watershed water pollution migration and transformation simulation method and system based on an MT-DHM model. The method comprises the following steps: collecting data from a watershed multi-source space, and performing data preprocessing to generate a hydrological response unit standard data set; performing vertical layering calculation based on the hydrological response unit standard data set to obtain land runoff vertical analysis data; performing land pollution migration simulation on the land runoff vertical analysis data, and performing real-time correction with an iteration step of 1h to obtain a dissolved / adsorbed load matrix; therefore, the present application solves the problems of insufficient data uniformity, low parameter calibration accuracy, and weak spatial distribution simulation capability of traditional models by constructing a full-process multi-source heterogeneous data fusion and pollution migration simulation mechanism, and improves the accuracy, stability, and management and control operability of watershed pollution load calculation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of migration simulation, and particularly relates to a watershed water pollution migration and transformation simulation method and system based on an MT-DHM model. BACKGROUND

[0002] Traditional watershed pollution migration models mostly rely on single physical processes or statistical methods, which are difficult to comprehensively capture the migration and transformation mechanism of pollutants in complex hydrological environments, resulting in large deviations in simulation results. Secondly, existing models are generally difficult to effectively fuse multi-source heterogeneous data, especially in terms of spatial heterogeneity and temporal dynamics, and cannot accurately reflect the spatio-temporal variation characteristics of different sub-regions and pollution sources in the watershed. In addition, traditional models often use fixed parameters or static calibration methods, which lack dynamic and multi-objective optimization capabilities, making it difficult to adapt to the nonlinear changes of watershed environmental conditions and pollution processes, resulting in limited prediction accuracy. Furthermore, current simulation methods generally lack detailed differentiation and coupling analysis for the different migration mechanisms of dissolved and adsorbed states of pollutants, which cannot fully reflect the conversion and accumulation processes of pollutants in different media, affecting the mechanism expression ability of the model. Finally, traditional systems lack detailed processing of the load transfer mechanism between multi-branch tree nodes in the multi-level simulation of the dilution, dispersion and attenuation processes of pollutants, and are difficult to accurately reflect the spatial propagation law of pollution load and key influence areas. SUMMARY

[0003] Therefore, it is necessary to provide a watershed water pollution migration and transformation simulation method and system based on an MT-DHM model to solve at least one of the above technical problems.

[0004] To achieve the above-mentioned purpose, a watershed water pollution migration and transformation simulation method based on an MT-DHM model, the method comprising the following steps:

[0005] Step S1: data collection is performed on the multi-source space of the watershed, and data preprocessing is performed to generate a standard data set of hydrological response units;

[0006] Step S2: vertical layering calculation is performed based on the standard data set of hydrological response units to obtain land runoff vertical analysis data; land pollution migration simulation is performed on the land runoff vertical analysis data, and real-time correction with an iteration step of 1h is performed to obtain a dissolved / adsorbed load matrix;

[0007] Step S3: push flow migration calculation is performed on the dissolved / adsorbed load matrix, and dispersion and dilution level joint simulation is performed to obtain watershed dilution model level data; automatic load transfer of multi-branch tree nodes is performed on the watershed dilution model level data to obtain an inter-node attenuation load matrix, wherein the inter-node attenuation load matrix includes an attenuation data sufficient area and an attenuation data deficient area;

[0008] Step S4: Perform mechanism modeling on the region with sufficient attenuation data to obtain pollutant load mechanism result data; perform statistical modeling on the region with scarce attenuation data to obtain pollutant load statistical result data; perform Bayesian fusion on the pollutant load mechanism result data and the pollutant load statistical result data to obtain the mixed load output matrix;

[0009] Step S5: Perform multi-objective calibration on the mixed load output matrix to obtain the pollution contribution rate distribution map; verify the accuracy of the pollution contribution rate distribution map to obtain the watershed water pollution simulation verification data; construct the Dianchi watershed sediment / nutrient qualification report using the watershed water pollution simulation verification data.

[0010] The beneficial effects of this invention are as follows: First, it acquires multi-source spatial data encompassing diverse and heterogeneous information, including DEM topographic data, land use data, soil parameters, and measured rainfall and runoff data. Second, it uses data cleaning techniques such as unified projection, time alignment, rasterization, and attribute reconstruction to form a standardized dataset of hydrological response units with consistent structure, laying a high-quality data foundation for subsequent analysis. Third, it constructs continuous hydrological profile data from vegetation interception and soil infiltration to groundwater runoff through vertical stratification calculation. Combined with measured parameters, it simulates and corrects pollutant migration in real time, achieving dynamic load calculation with a 1-hour step size and obtaining a dissolved / adsorbed pollution load matrix. Fourth, it models the spatial transport and attenuation process of pollutants as a multi-branch tree structure using a plug-flow migration and multi-path dispersion dilution model, effectively expressing the load transfer and zonal attenuation characteristics between upstream and downstream nodes. The attenuation data modeling is divided into mechanistic and statistical models. Bayesian fusion technology is used to jointly process structured and unstructured regional load data, significantly improving the robustness and data adaptability of the simulation results. In the model output stage, combining historical runoff, sediment, and nitrogen and phosphorus measured data, and employing multi-criteria model evaluation methods such as Nash efficiency coefficient and multi-objective calibration algorithms, the simulation results are spatially distributed and errors are optimized to ensure the high reliability of the pollution contribution rate distribution, thereby completing pollution risk early warning and compliance verification assessment. Ultimately, the compliance classification output of sediment and nutrient indicators at the watershed scale is achieved, providing direct data support and decision-making basis for environmental management and pollution control. Therefore, this invention, by constructing a full-process multi-source heterogeneous data fusion and pollution migration simulation mechanism, solves the problems of insufficient data uniformity, low parameter calibration accuracy, and weak spatial distribution simulation capability of traditional models, improving the accuracy, stability, and operability of watershed pollution load calculation and control.

[0011] Preferably, step S1 includes the following steps:

[0012] Step S11: Obtain hydrological dataset from hydrological monitoring station;

[0013] Step S12: Extract DEM data from the hydrological dataset at a resolution of 30m, perform watershed boundary analysis, and generate sub-watershed delineation maps;

[0014] Step S13: Extract river network data from the hydrological dataset according to the river section slope, and perform topological relationship parsing to obtain a multi-branch tree node encoding table;

[0015] Step S14: Divide the hydrological dataset into land use / soil type data according to 1km grids, and perform spatial overlay analysis to obtain the HRU cell attribute matrix;

[0016] Step S15: Perform Thiessen polygon spatial interpolation on the hydrological dataset at the hourly level to generate sub-basin surface rainfall sequences;

[0017] Step S16: Construct soil adsorption parameters from the hydrological dataset and perform laboratory calibration to obtain a watershed adsorption parameter table;

[0018] Step S17: Clean the data from the sub-basin division map, multi-branch tree node coding table, HRU unit attribute matrix, sub-basin surface rainfall sequence, and watershed adsorption parameter table, and format and unify them to obtain the standard dataset of hydrological response units.

[0019] This invention achieves standardized representation of watershed-scale hydrological response units at the data level through a systematic data hierarchical construction approach, providing high-quality structured input support for subsequent pollution load migration modeling and dynamic simulation. First, the method uses hydrological monitoring stations as the basic data source to acquire time-series hydrological observation data, ensuring the spatial representativeness and temporal continuity of the input data. In the DEM data extraction stage, 30m resolution high-precision digital elevation model data is used to identify topographic relief features and analyze river flow direction, thereby extracting refined sub-watershed boundaries and realizing spatial unit partitioning of the watershed. Subsequently, a river network backbone system is constructed based on river slope and spatial location. Topological relationship analysis generates a multi-branch tree node coding table that can be used for pollutant path simulation, effectively defining the direction of pollution migration and upstream-downstream hierarchical relationships. Land use and soil type information are spatially overlaid and analyzed after 1km rasterization to construct an HRU (Hydrological Response Unit) unit attribute matrix, allowing the hydrological response characteristics of different land-soil combinations to be reflected in the data structure. The areal rainfall sequence was generated based on hourly Thiessen polygon interpolation to ensure a complete representation of the spatiotemporal dynamics of rainfall. Simultaneously, soil adsorption parameters were calibrated and inverted using laboratory samples, transforming them into a structured watershed adsorption parameter table. Finally, all the aforementioned data were uniformly cleaned and integrated in terms of format, spatial coding, and temporal granularity to form a standard dataset of hydrological response units that is structurally complete, semantically clear, and granularly consistent. This dataset not only solves the problem of normalizing the representation of multi-source heterogeneous hydrological factors but also improves the input accuracy and spatial coverage integrity of pollution transport modeling, enhancing the applicability and interpretability of pollution process analysis models under complex terrain and diverse land cover conditions.

[0020] Preferably, step S2, which involves performing vertical stratification calculations based on the standard dataset of hydrological response units, includes:

[0021] Rainfall data was extracted from the standard dataset of hydrological response units to obtain watershed rainfall data; runoff process analysis was performed on the standard dataset of hydrological response units to obtain land watershed runoff data;

[0022] Vertical stratification calculations were performed on watershed rainfall data and land watershed runoff data to obtain vertical analysis data of land runoff. The vertical stratification calculations included: calculating the interception capacity of the vegetation interception layer based on a leaf area index (LAI) greater than 2.0 for both watershed rainfall data and land watershed runoff data to generate groundfall rainfall data; and performing infiltration analysis on the vegetation interception layer based on a saturation rate of the upper soil layer greater than 40% for both watershed rainfall data and land watershed runoff data to obtain the infiltration water component data of the land watershed.

[0023] Vertical data integration of groundfall rainfall data and land watershed infiltration data yields vertical analysis data of land runoff.

[0024] This invention extracts spatiotemporally consistent rainfall and surface runoff data from a standardized dataset of hydrological response units with a unified structure, and then conducts vertical structural modeling based on this data. In the vertical profile modeling process, the leaf area index (LAI) is first introduced as a discriminant factor for vegetation interception. When the LAI value is greater than 2.0, a quantitative estimation of the interception capacity is triggered, calculating the actual groundfall rainfall and thus eliminating the weakening effect of the vegetation layer on rainfall input. Based on this, according to the upper soil saturation, when the saturation rate is higher than 40%, an infiltration analysis mechanism is introduced to estimate the infiltration volume under the combined action of rainfall and runoff, finely dividing the surface runoff and infiltration components of vertical runoff. By vertically integrating the groundfall rainfall and infiltration component data, a land runoff vertical analysis dataset with process continuity and physical integrity is finally formed. This dataset not only covers the initial effects of rainfall, the interception behavior of hydrosurfaces, and the response processes of soil layers, but also establishes a multivariate data chain from rainfall-driven processes to infiltration transformation. It achieves vertical analysis and quantification of hydrological processes, providing particularly detailed input conditions and profile data support for modeling pollutant migration pathways, groundwater recharge, and soil moisture changes. This nested multi-source data fusion approach effectively enhances the model's ability to characterize vertical processes in complex surface systems and improves the numerical accuracy of the interaction mechanisms between pollutants and hydrological processes.

[0025] Preferably, step S2, which involves simulating land pollutant migration from the vertical analysis data of land runoff and performing real-time correction with an iteration step size of 1 hour, includes:

[0026] Obtain the soil permeability coefficient;

[0027] Based on soil permeability coefficient, the leaching rate of dissolved nitrogen or nitrate nitrogen is calculated from the vertical analysis data of land runoff to generate dissolved migration flux.

[0028] The amount of adsorbed phosphorus was calculated by analyzing the vertical analysis data of land runoff based on the amount of adsorbed phosphorus stripped when the slope of the soil shear force is greater than 5°.

[0029] The dissolved state migration flux and the adsorbed state migration flux are merged into adsorption levels and real-time correction is performed with an iteration step size of 1 hour to obtain the dissolved state / adsorbed state loading matrix.

[0030] This invention collects and inputs soil permeability coefficient data with spatial distribution attributes to parametrically model the hydraulic conduction characteristics of different soil layers in vertical land runoff analysis data. Based on these parameters, a leaching rate estimation method is used to quantitatively calculate the migration process of dissolved nitrogen or nitrate nitrogen in vertical water flow, forming dissolved migration flux expressed in time series form. On the other hand, for the migration mechanism of adsorbed phosphorus, a stripping mechanism criterion based on soil shear force and slope conditions is proposed. When the slope is greater than 5°, the stripping action function is activated, and the magnitude of phosphorus migration with particulate matter is estimated by a shear threshold analysis method, generating adsorbed migration data. Subsequently, the two types of migration data are merged according to adsorption hierarchy to construct a spatial-morphological coupled expression of pollution load. In the dynamic simulation stage, the iteration step is set to 1 hour, and the initial migration results are corrected in real time. The coefficients are gradually adjusted using the difference between historical load monitoring values ​​and simulation output to improve the time adaptability and response accuracy of load estimation. The resulting "dissolved / adsorbed load matrix" has the characteristics of clear hierarchy, balanced granularity, and complete time series in its data structure. It provides a high-precision and traceable data foundation for subsequent pollutant push calculations, load decay modeling, and watershed dilution analysis, and further promotes the transformation of the non-point source pollution modeling system from empirical estimation to a mechanism-driven and data-response parallel mode.

[0031] Preferably, step S3 includes the following steps:

[0032] Step S31: Perform plug flow migration calculation on the dissolved / adsorbed loading matrix, and calculate the flow velocity based on the roughness of 0.05 according to the Manning formula to obtain the inflow plug flow data;

[0033] Step S32: Perform a joint simulation of the dispersed dilution hierarchy based on the inflow propagation flow data to obtain the watershed dilution model hierarchy data;

[0034] Step S33: Automatically transfer the loads of the multi-branch tree nodes to the hierarchical data of the watershed dilution model to obtain the inter-node attenuation load matrix, which includes regions with sufficient attenuation data and regions with scarce attenuation data.

[0035] This invention utilizes the dissolved / adsorbed loading matrix as initial input and combines it with the boundary condition of a roughness coefficient of 0.05 in the Manning formula to perform parametric modeling and dynamic calculation of the flow velocity in the river cross section, obtaining highly timely inflow push flux data. This data not only reflects the ability of pollutants to move in the channel but also provides boundary driving variables for subsequent dilution simulations. In step S32, focusing on the physical diffusion, turbulence, and dispersion characteristics of pollutants during the flow process, physical parameters such as molecular diffusion coefficient, eddy viscosity correction factor, and river channel tortuosity are introduced to construct a joint model of dispersed dilution hierarchy, thereby forming a "watershed dilution model hierarchy data" with controllable spatial scale and distinct action mechanisms. This hierarchy data includes independent flux sets for each type of dilution process and an integrated comprehensive flux expression. Following step S33, the multi-branch tree node system of the river network topology in the basin is automatically identified and loaded. Based on the direction of flow and the hydrodynamic continuity between nodes, automatic pollution load transfer and attenuation estimation are performed. During this process, "attenuation data-sufficient areas" and "attenuation data-scarce areas" are automatically divided according to historical monitoring density and model data integrity. The former can enter the mechanism modeling path, and the latter can enter the statistical modeling path. This node attenuation structure constructs a high-dimensional accuracy support system for the spatiotemporal estimation of pollution distribution. The overall process not only ensures the physical rationality and data integrity of the pollutant migration link from inflow to channel to node, but also provides a traceable and decomposable data support system for subsequent multi-type model fusion, regional pollution classification diagnosis, and watershed-scale pollution control.

[0036] Preferably, step S32 includes the following steps:

[0037] Step S321: Perform diffusion coefficient analysis on the inflow propulsion data. ―5 m 2 The molecular diffusion dilution data were obtained by calculating the molecular diffusion dilution data using Fick's law for / s.

[0038] Step S322: Correct the inflow propulsion flux data for eddy viscosity coefficient with a Reynolds number greater than 4000 to obtain turbulent diffusion dilution data;

[0039] Step S323: Perform shear flow discretization analysis on the inflow propulsion data with a channel curvature greater than 1.2 to obtain dilution data due to dispersion effect;

[0040] Step S324: Integrate the molecular diffusion dilution data, turbulent diffusion dilution data, and dispersion dilution data to obtain the watershed dilution model hierarchical data.

[0041] This invention performs molecular diffusion calculations on inflow propulsion data using a set diffusion coefficient, generating microscopic diffusion dilution data based on molecular motion. This data accurately characterizes the diffusion and expansion features of pollutants at the molecular level. Step S322, for high Reynolds number (Re>4000) flow conditions, corrects the turbulent diffusion effect in the flow field through an eddy viscosity coefficient correction mechanism, generating turbulent diffusion dilution data reflecting the enhanced mixing effect of turbulence, effectively compensating for the limitations of molecular diffusion under intense flow conditions. Step S323 uses a river channel meandering threshold (greater than 1.2) as a criterion to conduct shear flow discrete analysis, quantitatively assessing the impact of river meandering on fluid shear force, thereby forming dilution data reflecting dispersion. This process fully considers the local influence of complex hydraulic topography on pollutant distribution. Step S324 uses a multi-source data fusion algorithm to structurally integrate the above three types of dilution data, generating unified watershed dilution model hierarchical data. This dataset takes into account the physical nature and temporal and spatial dynamics of different diffusion processes, supporting subsequent watershed water quality evolution simulation and pollution risk assessment. The overall technology enables multi-dimensional and multi-scale analysis of pollutant diffusion processes within the watershed, enhancing the model's responsiveness to the complexity of the actual water environment and the reliability of its predictions.

[0042] Preferably, step S4 includes the following steps:

[0043] Step S41: Perform mechanism modeling on the region with sufficient attenuation data to obtain pollutant load mechanism result data;

[0044] Step S42: Select areas with a forest area ratio greater than 70% for multiple regression model construction in areas with scarce attenuation data, and estimate forest load to obtain pollutant load statistical results;

[0045] Step S43: Perform Bayesian fusion of pollutant load mechanism results data and pollutant load statistical results data to obtain a mixed load output matrix.

[0046] This invention constructs a mathematical model based on the pollutant generation, migration, and transformation mechanisms in areas with abundant attenuation data, relying on complete water quality, hydrological, and environmental variable data. By parametrically describing key biochemical processes such as nitrification and denitrification, it achieves physical mechanism analysis of dynamic changes in pollutant load, generating pollutant load mechanism results data with high spatiotemporal resolution and physical interpretability. Step S42, for data-scarce areas, selects regions with forest coverage exceeding 70% as samples, applies multiple regression analysis to construct a statistical prediction model, and combines land use, meteorological, and existing load observation data to achieve quantitative estimation of forest pollution load, obtaining pollutant load statistical results data driven by empirical data. Step S43, through a Bayesian fusion algorithm, integrates the data output from the mechanistic model with the results of the statistical model using probability weighting, fully considering the credibility and uncertainty of each model, achieving optimal fusion output of pollution load information, and generating a mixed load output matrix. This matrix possesses both the physical rationality of the mechanistic model and the empirical compensation capability of the statistical model, overcoming the limitations of single modeling methods under different regional data conditions. The overall process effectively enhances the spatial continuity and temporal dynamics of pollution load estimation, providing a high-quality input data foundation for subsequent watershed pollutant transport simulation and risk assessment.

[0047] Preferably, step S41 includes the following steps:

[0048] Step S411: Analyze nitrification at temperatures above 15℃ based on sufficient attenuation data to obtain ammonia nitrogen conversion.

[0049] Step S412: Based on the sufficient attenuation data, the denitrification effect was analyzed for DO less than 2 mg / L, and first-order kinetics were used to obtain nitrogen removal data;

[0050] Step S413: Compile the ammonia nitrogen conversion and denitrification data according to a monitoring point density greater than 1 / 100km. 2 Mechanism modeling was performed to obtain data on pollutant load mechanism results.

[0051] This invention, under environmental conditions with temperatures above 15°C, meticulously calculates the nitrification process of ammonia nitrogen based on on-site monitoring data and experimental calibration parameters. Combined with a temperature-dependent nitrification rate model, it generates time-series data reflecting the conversion of ammonia nitrogen to nitrate nitrogen. This data accurately describes the kinetic characteristics of the nitrification stage in the nitrogen cycle. Step S412 focuses on anoxic environments with dissolved oxygen (DO) concentrations below 2 mg / L, using a first-order kinetic model to quantitatively simulate the denitrification process. Based on on-site DO monitoring data and denitrification rate parameters, it derives the dynamic changes in nitrogen removal, forming a spatiotemporal estimate of nitrogen loss. Step S413 divides the above nitrification conversion and nitrogen removal data according to the spatial density of monitoring points (greater than 1 per 100 km).2 Spatial interpolation and mechanistic modeling are performed, and spatial statistical methods are used to estimate the continuous spatial distribution of pollutant loads, generating high-precision pollutant load mechanism data. This data reflects the dynamic evolution of key nitrogen transformation processes over time and demonstrates the impact of environmental heterogeneity on pollution loads spatially, enhancing the physical interpretability and predictive power of pollutant load models and providing a reliable data foundation for watershed pollution control and management.

[0052] Preferably, step S5 includes the following steps:

[0053] Step S51: Perform multi-objective calibration on the mixed load output matrix to obtain multi-objective calibration data; render the multi-objective calibration data into a heat map to obtain a pollution contribution rate distribution map;

[0054] Step S52: Verify the accuracy of the pollution contribution rate distribution map to obtain watershed water pollution simulation verification data;

[0055] Step S53: Construct a qualified report for sediment / nutrients in the Dianchi Lake basin using the simulated verification data of water pollution in the basin.

[0056] This invention employs a multi-objective calibration process based on historical monitoring and simulation data, utilizing a multi-index evaluation system to jointly correct the spatial distribution and temporal evolution of pollutant loads. This yields multi-objective calibration data that reflects the contribution characteristics of various pollutants within the watershed. After heatmap rendering, this data visually presents the spatial heterogeneity of pollution contribution rates, forming a pollution contribution rate distribution map, providing a quantitative basis for watershed pollution source identification and management strategies. In step S52, the pollution contribution rate distribution map is validated against actual monitoring data. Statistical indicators such as correlation coefficient and root mean square error are used to assess the accuracy and reliability of the model simulation, generating watershed water pollution simulation validation data to ensure consistency between the model output and actual water quality conditions. Step S53 systematically organizes the validated simulation data and, based on national and local water quality standards, constructs a sediment and nutrient compliance report for the Dianchi Lake watershed, reflecting the watershed's water quality status and treatment effectiveness. This report not only provides scientific decision-making support for environmental management departments but also provides data support for the formulation of subsequent pollutant load control and water quality improvement measures. The entire process, from data collection and model calibration to verification and report construction, forms a complete closed loop, effectively improving the accuracy and application value of watershed water pollution simulation and promoting the scientific management and sustainable development of the watershed environment.

[0057] This specification provides a watershed water pollution migration and transformation simulation system based on the MT-DHM model, used to execute the aforementioned watershed water pollution migration and transformation simulation method based on the MT-DHM model. This watershed water pollution migration and transformation simulation system based on the MT-DHM model includes:

[0058] The watershed data acquisition and standardization module is used to acquire data from multiple sources in the watershed and perform data preprocessing to generate standard datasets for hydrological response units.

[0059] The vertical runoff and migration correction module is used to perform vertical stratification calculations based on the standard dataset of hydrological response units to obtain vertical analysis data of land runoff; it simulates the migration of land pollutants from the vertical analysis data of land runoff and performs real-time correction with an iteration step of 1 hour to obtain the dissolved / adsorbed load matrix.

[0060] The water system migration and dilution hierarchy simulation module is used to perform push-flow migration calculations on the dissolved / adsorbed load matrix and to perform joint simulation of the dispersed dilution hierarchy to obtain watershed dilution model hierarchy data. The watershed dilution model hierarchy data is then automatically transferred to the nodes of a multi-branch tree to obtain the inter-node attenuation load matrix, which includes regions with sufficient attenuation data and regions with scarce attenuation data.

[0061] The mechanism / statistical modeling and Bayesian fusion module is used to perform mechanism modeling in areas with sufficient attenuation data to obtain pollutant load mechanism result data; to perform statistical modeling in areas with scarce attenuation data to obtain pollutant load statistical result data; and to perform Bayesian fusion of pollutant load mechanism result data and pollutant load statistical result data to obtain a mixed load output matrix.

[0062] The contribution rate calibration and qualified report generation module is used to perform multi-objective calibration on the mixed load output matrix to obtain a pollution contribution rate distribution map; to verify the accuracy of the pollution contribution rate distribution map to obtain watershed water pollution simulation verification data; and to construct a qualified report for sediment / nutrients in the Dianchi Lake watershed using the watershed water pollution simulation verification data. Attached Figure Description

[0063] Figure 1 This is a schematic diagram of the steps in a watershed water pollution migration and transformation simulation method based on the MT-DHM model.

[0064] Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S3.

[0065] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0066] The technical method of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this invention.

[0067] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0068] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0069] To achieve the above objectives, please refer to Figures 1 to 2 A watershed water pollution migration and transformation simulation method based on the MT-DHM model, the method comprising the following steps:

[0070] Step S1: Collect data from multiple sources in the watershed and preprocess the data to generate a standard dataset of hydrological response units;

[0071] Step S2: Perform vertical stratification calculations based on the standard dataset of hydrological response units to obtain vertical analysis data of land runoff; simulate the migration of land pollutants based on the vertical analysis data of land runoff, and perform real-time correction with an iteration step size of 1 hour to obtain the dissolved / adsorbed load matrix;

[0072] Step S3: Perform push-flow migration calculations on the dissolved / adsorbed load matrix and conduct joint simulation of the dispersion dilution hierarchy to obtain the watershed dilution model hierarchy data; perform automatic load transfer on the watershed dilution model hierarchy data through multi-branch tree nodes to obtain the inter-node attenuation load matrix, wherein the inter-node attenuation load matrix includes regions with sufficient attenuation data and regions with scarce attenuation data.

[0073] Step S4: Perform mechanism modeling on the region with sufficient attenuation data to obtain pollutant load mechanism result data; perform statistical modeling on the region with scarce attenuation data to obtain pollutant load statistical result data; perform Bayesian fusion on the pollutant load mechanism result data and the pollutant load statistical result data to obtain the mixed load output matrix;

[0074] Step S5: Perform multi-objective calibration on the mixed load output matrix to obtain the pollution contribution rate distribution map; verify the accuracy of the pollution contribution rate distribution map to obtain the watershed water pollution simulation verification data; construct the Dianchi watershed sediment / nutrient qualification report using the watershed water pollution simulation verification data.

[0075] In this embodiment of the invention, reference is made to Figure 1 The diagram shows a flowchart illustrating the steps of a watershed water pollution migration and transformation simulation method based on the MT-DHM model according to the present invention. In this example, the watershed water pollution migration and transformation simulation method based on the MT-DHM model includes the following steps:

[0076] Step S1: Collect data from multiple sources in the watershed and preprocess the data to generate a standard dataset of hydrological response units;

[0077] In this embodiment of the invention, representative remote sensing images (such as Sentinel-2 or Landsat series) are preferentially selected as the basic source of surface information for data acquisition. Simultaneously, heterogeneous data of various types are acquired, including meteorological station rainfall data, hydrological station flow data, soil attribute databases, land use status maps, DEM elevation data, and administrative division vector boundary information. These data typically suffer from inconsistencies in spatial scale, temporal resolution, and observation accuracy, thus requiring preliminary standardization through resampling, interpolation, coordinate system unification, and spatial clipping. Next, based on topographic analysis and flow-discharge extrapolation methods, the DEM data undergoes sub-basin division and river network extraction. Combined with land use classification layers and soil type data, partitioning is performed to construct basic hydrological response units (HRUs) associated with geomorphic units, hydrological cycles, and pollutant transport processes. To ensure the usability of the data in subsequent vertical simulations and pollution transfer processes, missing or abnormally observed data points are repaired and corrected using a combination of historical time-series trend regression and neighborhood spatial interpolation. Meanwhile, to ensure the consistency and operability of the dataset, all input factors (such as slope, surface roughness, soil permeability, vegetation index, and surface runoff coefficient) were normalized according to preset standardization rules, and the processing results were mapped to the hydrological response unit grid structure, ultimately forming a data foundation framework with consistent spatial attributes and a unified temporal dimension, suitable for downstream simulation analysis. The dataset constructed in this step is not only a key foundation supporting subsequent vertical runoff stratification calculations and pollution migration modeling, but also an important prerequisite for ensuring the calculation accuracy and regional representativeness of the entire simulation system.

[0078] Step S2: Perform vertical stratification calculations based on the standard dataset of hydrological response units to obtain vertical analysis data of land runoff; simulate the migration of land pollutants based on the vertical analysis data of land runoff, and perform real-time correction with an iteration step size of 1 hour to obtain the dissolved / adsorbed load matrix;

[0079] In this embodiment of the invention, within the hydrological response unit grid, slope data derived from the DEM, land use type, and soil profile parameters (such as soil layer thickness, organic matter content, volumetric water content, saturated hydraulic conductivity, etc.) are combined to perform vertical stratification calculations based on the surface-soil-subsurface three-layer structure. Hydraulic parameters between each layer are extracted from the national soil database and land use classification table using lookup table methods and spatial interpolation techniques, and mapped to each response unit grid. After the vertical profile is established, the modified Green-Ampt model or multi-level Richards equations are used to dynamically simulate the downward infiltration of surface precipitation, soil water storage, and deep seepage processes, obtaining water flux data between layers per unit time step. Subsequently, based on the established water flux, pollutant input source data are introduced, including surface pollutant deposition flux, fertilizer application statistics, livestock and poultry farming emission coefficients, and the intensity of non-point source pollution from residential areas. These pollutants are then categorized into dissolved and adsorbed states according to their morphological characteristics. In the simulation of dissolved pollutant migration, an Advancement-Dispersion transport model combined with real-time soil moisture data was used to calculate the convective diffusion behavior of pollutants in the vertical profile. For the adsorbed state, the Freundlich isotherm adsorption function was introduced to characterize the proportion of pollutants adsorbed onto the solid phase by soil particles, and this was used to dynamically adjust the distribution relationship of pollutants between the liquid and solid phases. The entire simulation process was executed iteratively in 1-hour increments, with real-time corrections using observational data or historical simulation output data to ensure the stability and rationality of the migration model under actual boundary conditions. The final output is a dissolved / adsorbed load matrix reflecting the multi-form transformation and enrichment process of pollutants in the vertical land system. This matrix serves as an intermediate result for pollutant migration, providing high-resolution, highly dynamic input data support for subsequent water system migration and dilution simulations.

[0080] Step S3: Perform push-flow migration calculations on the dissolved / adsorbed load matrix and conduct joint simulation of the dispersion dilution hierarchy to obtain the watershed dilution model hierarchy data; perform automatic load transfer on the watershed dilution model hierarchy data through multi-branch tree nodes to obtain the inter-node attenuation load matrix, wherein the inter-node attenuation load matrix includes regions with sufficient attenuation data and regions with scarce attenuation data.

[0081] In this embodiment of the invention, a flow-discharge map is constructed based on the spatial grid structure of hydrological response units and a DEM (Digital Elevation Model). Combined with the river network structure of the basin and measured or simulated daily runoff data, the confluence paths and hydraulic connections of each tributary unit are determined. Subsequently, the load matrix output from the previous step is mapped to the corresponding hydrological unit. Pollutant migration with the flow is calculated using the load transfer rate per unit length of the river channel and hydrodynamic parameters (such as cross-sectional velocity, water depth, and hydraulic gradient). The pollutant propagation model employs a one-dimensional or quasi-two-dimensional water quality transport equation (such as the Advance-Dispersion Equation, ADE), evolving the load changes step-by-step with hourly time steps. To simulate the dilution effect of the load during transport, process variables such as tributary inflow, rainfall replenishment, and evapotranspiration loss are further introduced. By constructing a decentralized dilution hierarchy structure based on node convergence rules, the pollutant concentrations of different levels of confluence units are updated layer by layer, thereby forming a multi-level, stratified pollution dilution simulation result within the basin. Building upon this foundation, to achieve automated analysis of pollution load propagation and attenuation in complex watershed networks, a multi-branch tree structure is used to organize the nodes of the watershed network. Each node represents a river segment or confluence point, and the pollution load transfer between nodes is calculated comprehensively based on the total upstream load, velocity factor, node residence time, and local source input. When constructing the attenuation relationships between nodes, empirical attenuation models (such as exponential or power function attenuation models) are used to fit the attenuation patterns of pollutants between nodes, dividing the data into data-sufficient and data-scarce regions, and generating a complete pollution load attenuation matrix between nodes. This provides a clear and dynamically adjustable input foundation for subsequent mechanistic and statistical modeling. This matrix records the pollutant transfer amount, attenuation rate, and time delay characteristics between nodes in hourly time series format, serving as the core carrier for subsequent multi-source modeling and fusion calculations.

[0082] Step S4: Perform mechanism modeling on the region with sufficient attenuation data to obtain pollutant load mechanism result data; perform statistical modeling on the region with scarce attenuation data to obtain pollutant load statistical result data; perform Bayesian fusion on the pollutant load mechanism result data and the pollutant load statistical result data to obtain the mixed load output matrix;

[0083] In this embodiment of the invention, mechanistic modeling and statistical modeling methods are used to construct pollutant load evolution models, respectively. The results of both methods are then integrated using a Bayesian fusion approach to construct a hybrid load output matrix at the data level. In areas with sufficient attenuation data, multidimensional variables such as time-series pollutant concentration, flow velocity, water temperature, and water retention time are extracted by collecting historical observation data, measured water quality and quantity data, and process simulation results. Based on hydrodynamics and water transport theory, a one-dimensional reaction-transport model or a retention zone coupling model is used to establish the transport and attenuation relationship of pollutants. Such mechanistic models often employ first-type kinetic reaction equations or multi-component solute transformation mechanisms to describe the physicochemical processes of pollutant degradation, transformation, and sedimentation in water bodies, while simultaneously coupling river cross-sections, flow paths, and node topological features in spatial structure. Model parameters such as reaction rate coefficients, sedimentation rates, and partition volume ratios are determined through inversion or sensitivity analysis of measured data, and pollutant attenuation is dynamically simulated on an hourly scale, outputting high spatiotemporal resolution pollutant load response data. In areas with scarce attenuation data, due to the lack of sufficient process data, a statistical modeling approach is necessary. This involves incorporating historical data from spatially neighboring nodes, regional land use structure, hydrological response unit types, and upstream load input data to construct pollutant load estimation models based on linear regression, principal component regression, or random forest regression. During data processing, normalization transformation, feature importance screening, and cross-validation strategies are employed to enhance model robustness and generalization ability, ultimately outputting the pollution load estimates and their uncertainty ranges for the corresponding nodes. To achieve synergistic fusion of mechanistic and statistical model results, a Bayesian fusion mechanism is introduced. Using the prior probability distribution of pollutant load as a foundation, and combining the conditional likelihood function of the mechanistic model with the confidence prediction interval of the statistical model, a posterior probability distribution is constructed. The pollutant mixed load output matrix is ​​obtained in the form of maximum a posteriori estimation. This matrix comprehensively records the pollutant load and its confidence interval for each node at a given time scale, providing a quantitative data input basis for subsequent pollution contribution rate calibration and simulation accuracy verification.

[0084] Step S5: Perform multi-objective calibration on the mixed load output matrix to obtain the pollution contribution rate distribution map; verify the accuracy of the pollution contribution rate distribution map to obtain the watershed water pollution simulation verification data; construct the Dianchi watershed sediment / nutrient qualification report using the watershed water pollution simulation verification data.

[0085] In this embodiment of the invention, a mixed load output matrix is ​​used as input. This matrix contains information on pollutant concentrations, speciation ratios, and migration states at different time scales for each node within the watershed. Combining measured water quality data (such as total phosphorus, total nitrogen, COD, TSS, etc.) and corresponding hydrological data (such as flow rate, velocity, water level, etc.), a multi-objective calibration function is constructed with the objectives of minimizing pollutant concentration deviation, matching the pollution process response time, and ensuring spatial load distribution consistency. To achieve a balance between calibration accuracy and stability, a multi-objective genetic algorithm (MOGA) or particle swarm optimization (PSO) is used for global parameter optimization. Simultaneously, sensitivity analysis is performed on key parameters in the model structure (such as source strength coefficient, attenuation coefficient, sedimentation rate, and dilution factor), and constraint boundaries are set during the calibration process to avoid model overfitting. After calibration, a retrospective analysis of the pollutant sources at each response unit or sub-watershed node is conducted. Based on the spatiotemporal load contribution relationship, the spatial distribution ratio of pollution sources is calculated, and finally, a pollution contribution rate distribution map is output. Technically, this map uses spatial interpolation methods (such as Kriging or IDW) to continuously represent the contribution rate data of discrete nodes, and overlays administrative division, land use, and water system layers to enhance geographic identification. In the accuracy verification stage, measured water quality data from different calibration periods are selected as independent validation sets. The fit between the model output and the observed data is evaluated by calculating indicators such as the Nash efficiency coefficient (NSE), root mean square error (RMSE), and relative error (RE). After the verification results meet the accuracy requirements, using the pollution contribution rate distribution map as the core reference, and combining the Dianchi Lake basin water quality standards, total emission control indicators, historical pollution evolution trends, and key monitoring data, a sediment and nutrient compliance report is constructed. The report quantitatively lists the compliance status of each sub-region, the pollution source structure, potential exceedance risks, and recommended control strategies, forming a pollution simulation output with quantitative support and spatial interpretability.

[0086] Preferably, step S1 includes the following steps:

[0087] Step S11: Obtain hydrological dataset from hydrological monitoring station;

[0088] Step S12: Extract DEM data from the hydrological dataset at a resolution of 30m, perform watershed boundary analysis, and generate sub-watershed delineation maps;

[0089] Step S13: Extract river network data from the hydrological dataset according to the river section slope, and perform topological relationship parsing to obtain a multi-branch tree node encoding table;

[0090] Step S14: Divide the hydrological dataset into land use / soil type data according to 1km grids, and perform spatial overlay analysis to obtain the HRU cell attribute matrix;

[0091] Step S15: Perform Thiessen polygon spatial interpolation on the hydrological dataset at the hourly level to generate sub-basin surface rainfall sequences;

[0092] Step S16: Construct soil adsorption parameters from the hydrological dataset and perform laboratory calibration to obtain a watershed adsorption parameter table;

[0093] Step S17: Clean the data from the sub-basin division map, multi-branch tree node coding table, HRU unit attribute matrix, sub-basin surface rainfall sequence, and watershed adsorption parameter table, and format and unify them to obtain the standard dataset of hydrological response units.

[0094] In this embodiment of the invention, hourly or daily-scale hydrological datasets, including time-series data such as rainfall, evaporation, runoff, flow velocity, and water level, are acquired by hydrological monitoring stations deployed at the inlets and outlets of rivers, tributaries, and reservoirs, and matched with corresponding spatial location information. Step S12 uses a digital elevation model (DEM) with a spatial resolution of 30m, employs a flow direction extraction algorithm (such as the D8 algorithm) and flow accumulation analysis to determine the flow direction and catchment area, and identifies the watershed boundary by setting a confluence threshold, thus generating a sub-watershed division map. Step S13 uses the river segment slope derived from the DEM as an index to extract the river network system, and establishes a directed graph structure to encode the numbering and topological relationships of multi-branch tree nodes based on the upstream-downstream hydraulic connection relationship, generating a multi-branch tree node encoding table containing attribute fields such as node number, parent-child node pairs, river segment length, and slope. Step S14 uses a 1km raster as the basic analysis unit to perform spatial resampling and overlay analysis on land use data (such as remote sensing classification layers) and soil type data (such as the National Soil Database). Based on the land use-soil type combination logic, HRU (Hydrological Response Unit) units are constructed, forming an HRU unit attribute matrix. Each row in this matrix corresponds to a raster unit and includes key parameters such as soil type, land use type, slope grade, and surface permeability. Step S15 performs spatial interpolation on the point-based rainfall data, constructing a Voronoi diagram to map the station rainfall data to the corresponding polygon control area. It also generates hourly areal rainfall time series for each sub-basin, reflecting the spatial distribution continuity of rainfall. Step S16 conducts adsorption experiments based on field-collected soil samples. By constructing Freundlich or Langmuir isotherm adsorption models and combining the experimental fitting results, parameters such as adsorption capacity per unit mass and adsorption constant are extracted to form a structured watershed adsorption parameter table. Finally, in step S17, the sub-basin division map, multi-branch tree coding table, HRU attribute matrix, sub-basin rainfall sequence and adsorption parameter table are spatially registered in a unified coordinate system. Data cleaning techniques are used to handle missing and outlier values, the field structure is standardized, the coding method is formatted, and a unified data framework is used to output a standard dataset of hydrological response units, which serves as the core data foundation to support watershed pollutant migration modeling and multi-scale hydrological simulation analysis.

[0095] Preferably, step S2, which involves performing vertical stratification calculations based on the standard dataset of hydrological response units, includes:

[0096] Rainfall data was extracted from the standard dataset of hydrological response units to obtain watershed rainfall data; runoff process analysis was performed on the standard dataset of hydrological response units to obtain land watershed runoff data;

[0097] Vertical stratification calculations were performed on watershed rainfall data and land watershed runoff data to obtain vertical analysis data of land runoff. The vertical stratification calculations included: calculating the interception capacity of the vegetation interception layer based on a leaf area index (LAI) greater than 2.0 for both watershed rainfall data and land watershed runoff data to generate groundfall rainfall data; and performing infiltration analysis on the vegetation interception layer based on a saturation rate of the upper soil layer greater than 40% for both watershed rainfall data and land watershed runoff data to obtain the infiltration water component data of the land watershed.

[0098] Vertical data integration of groundfall rainfall data and land watershed infiltration data yields vertical analysis data of land runoff.

[0099] In this embodiment of the invention, the spatial boundary of a hydrological response unit is used as a mask to extract the corresponding sub-basin surface rainfall sequence, generating "basin rainfall data" with a time index. Simultaneously, within the same response unit, a runoff generation model driven by DEM, land use, slope, and HRU parameters (such as the SCS-CN method, Green-Ampt, or a simplified TOPMODEL) is invoked to simulate the surface runoff process, generating "land watershed runoff data." Subsequently, the vertical stratification calculation stage begins, where the system constructs a vertical profile with a three-layer structure: "vegetation interception layer—soil surface layer—deep soil layer." In vegetation interception modeling, the Leaf Area Index (LAI) obtained from remote sensing is used based on the vegetation type of each response unit, and units with an LAI greater than 2.0 are selected to simulate significant interception processes. The interception capacity per unit time is calculated based on the empirical relationship between LAI and rainfall intensity using tables or empirical formulas (such as the Gash model or Rutter model), and then the corresponding interception amount is subtracted from the original rainfall data to obtain "groundfall rainfall data," i.e., the net rainfall reaching the soil surface. In the modeling of the infiltration process, soil units with a saturated water content, initial water content, and permeability coefficient of different soil layers in the soil property database are selected, and simplified infiltration estimation is performed using the Green-Ampt or Richard equations to simulate the infiltration behavior of "falling rainfall" into the surface soil per unit time, outputting "infiltration water component data". Finally, the above "falling rainfall data" and "land watershed infiltration water component data" are aligned one by one in spatial dimension and time index, and vertical path integration is performed to construct complete "land runoff vertical analysis data". This data uses response units as the basic structure and records the distribution of net rainfall, infiltrate, and actual infiltration per unit area at each time step, providing a vertical input framework and boundary constraints for subsequent modeling of terrestrial pollutant migration processes.

[0100] Preferably, step S2, which involves simulating land pollutant migration from the vertical analysis data of land runoff and performing real-time correction with an iteration step size of 1 hour, includes:

[0101] Obtain the soil permeability coefficient;

[0102] Based on soil permeability coefficient, the leaching rate of dissolved nitrogen or nitrate nitrogen is calculated from the vertical analysis data of land runoff to generate dissolved migration flux.

[0103] The amount of adsorbed phosphorus was calculated by analyzing the vertical analysis data of land runoff based on the amount of adsorbed phosphorus stripped when the slope of the soil shear force is greater than 5°.

[0104] The dissolved state migration flux and the adsorbed state migration flux are merged into adsorption levels and real-time correction is performed with an iteration step size of 1 hour to obtain the dissolved state / adsorbed state loading matrix.

[0105] In this embodiment of the invention, a quantitative parameter-driven vertical water quality migration model is used to simulate and dynamically integrate the migration behavior of nitrogen and phosphorus pollutants during land runoff. First, statistical analysis is performed on the saturated hydraulic conductivity (Ksat) and effective porosity of different soil types in the soil dataset to obtain soil permeability coefficient data for corresponding spatial units. This data is then mapped according to hydrological response units to form a spatially matched permeability parameter grid. In the simulation of dissolved nitrogen or nitrate nitrogen migration, previously generated vertical analysis data of land runoff is used. Combining the infiltration rate per unit time with the soil permeability coefficient, a first-order momentum conservation and mass conservation coupled model is employed to estimate the leaching rate of nitrogen carried by water migrating downwards. The core calculation basis is the coupling relationship between hydraulic migration flux and nitrogen concentration gradient, and a "dissolved migration flux" data matrix is ​​generated with an hourly time step. For the migration process of adsorbed pollutants, spatial units with a slope greater than 5° are selected. Slope distribution is extracted from the existing HRU unit attribute matrix. An empirical slope erosion model (such as the Revised Universal Soil Loss Equation, RUSLE) is constructed based on the relationship between soil particle attached phosphorus content and the critical shear force for stripping. Soil erosion modulus and erosion response coefficient are introduced for adjustment to estimate the amount of adsorbed phosphorus stripped per unit area, generating an "adsorbed migration amount" dataset. In the morphology fusion stage, "dissolved migration flux" and "adsorbed migration amount" are aligned and weighted according to spatial units to construct a three-dimensional "adsorption hierarchy merging matrix." This matrix retains hourly granularity in the time dimension and records the dynamic transfer values ​​of each pollutant form within a specific unit. To ensure consistency between simulation results and actual monitoring data, a real-time correction mechanism at the hourly scale is further introduced. Kalman filtering or a least-squares dynamic regression algorithm based on the bias function is used to update the predicted values ​​of the merged matrix hourly, ultimately outputting a "dissolved / adsorbed load matrix". This matrix comprehensively expresses the hydrodynamic-chemical dynamic coupling migration law of pollutants in the vertical path of the soil, providing fine-grained input data for subsequent plugging evolution and dilution modeling.

[0106] As an example of the present invention, reference is made to... Figure 2 As shown, step S3 in this example includes:

[0107] Step S31: Perform plug flow migration calculation on the dissolved / adsorbed loading matrix, and calculate the flow velocity based on the roughness of 0.05 according to the Manning formula to obtain the inflow plug flow data;

[0108] Step S32: Perform a joint simulation of the dispersed dilution hierarchy based on the inflow propagation flow data to obtain the watershed dilution model hierarchy data;

[0109] Step S33: Automatically transfer the loads of the multi-branch tree nodes to the hierarchical data of the watershed dilution model to obtain the inter-node attenuation load matrix, which includes regions with sufficient attenuation data and regions with scarce attenuation data.

[0110] In this embodiment of the invention, during the flow migration calculation stage, the dissolved / adsorbed load matrix is ​​first mapped to river segment spatial units, allocating the pollutant load of each HRU unit or sub-basin to the corresponding river segment inlet location, thus establishing a spatial location index for the pollutant input points. For velocity calculation, the Manning formula is used. In this model, n = 0.05, and the hydraulic radius R and river slope S are obtained from the DEM and river segment parameters. The flow velocity per unit time for each river segment is calculated using this formula, and the volumetric flow rate is estimated by combining the cross-sectional width and depth, thus obtaining the "inflow thrust data" per unit time. In the joint simulation of the dispersed dilution hierarchy, the inflow thrust data is used as input, combined with parameters such as the change in flow along the river, river temperature, and background water quality concentration. A pollutant dilution simulation model is established using a multi-scale diffusion-convection equation. The model considers the velocity difference and concentration gradient change at the confluence of the main channel and tributaries, and introduces parameters such as diffusion coefficient and distribution coefficient to generate "basin dilution model hierarchy data" at different spatial scales. This data retains fields such as river segment code, pollutant concentration evolution curve over time, and dilution attenuation factor in its structure. Subsequently, under the multi-branch tree network node hierarchy, the system performs path tracking on the outlet load data of each river segment and constructs an automatic load transfer mechanism, that is, the output of the upstream node is used as the input of the downstream node, and an "inter-node attenuation load matrix" is formed through accumulation and reduction mechanisms. In this process, for areas with complete data (i.e., supported by measured or historical verification data), a mechanism model based on physical parameters is used to calculate the load attenuation value; while in areas with scarce data, they are marked as incomplete attenuation areas, and statistical interpolation or missing data compensation strategies are introduced to preserve the continuity of load transmission between nodes, thereby ensuring that the "inter-node attenuation load matrix" meets the modeling requirements for subsequent pollutant load allocation and contribution rate analysis in terms of structure and data consistency.

[0111] Preferably, step S32 includes the following steps:

[0112] Step S321: Perform diffusion coefficient analysis on the inflow propulsion data. ―5 m 2 The molecular diffusion dilution data were obtained by calculating the molecular diffusion dilution data using Fick's law for / s.

[0113] Step S322: Correct the inflow propulsion flux data for eddy viscosity coefficient with a Reynolds number greater than 4000 to obtain turbulent diffusion dilution data;

[0114] Step S323: Perform shear flow discretization analysis on the inflow propulsion data with a channel curvature greater than 1.2 to obtain dilution data due to dispersion effect;

[0115] Step S324: Integrate the molecular diffusion dilution data, turbulent diffusion dilution data, and dispersion dilution data to obtain the watershed dilution model hierarchical data.

[0116] In this embodiment of the invention, a concentration gradient field is first constructed based on the inflow flux data, and then Fick's first law is applied. The flux of pollutant concentration at the microscopic level of the fluid is calculated as a function of space, where D represents the molecular diffusion coefficient (unit: m). 2 / s), this parameter is obtained from empirical tables or experimental databases based on pollutant type and water temperature conditions, and the calculation results generate "molecular diffusion dilution data", recording the microscopic concentration diffusion intensity at each river segment location. In step S322, the Reynolds number of each river segment is calculated. The fully turbulent region with Re greater than 4000 was identified, and an eddy viscosity coefficient was introduced based on the principle of turbulent viscosity for correction. An improved Elder formula or a similar eddy diffusion model D was used. t =αuh, where u is the flow velocity, h is the water depth, and α is the empirical coefficient of turbulent diffusion. This estimates the macroscopic dilution behavior under turbulent conditions, generating "turbulent diffusion dilution data." Step S323 focuses on the geometric changes of the river channel. Based on the river section curvature index extracted from the Geographic Information System (GIS), areas with curvature greater than 1.2 are selected, and a shear flow field profile is established. The shear effect caused by the velocity gradient between different longitudinal water layers is calculated in the two-dimensional flow field. Then, the longitudinal diffusion coefficient is calculated using the Fischer dispersion model. Where H represents water depth and K is the tortuosity correction factor, the discrete migration process of pollutants caused by shear flow is simulated, and "diffusion dilution data" is output. The above three types of dilution data are fused by attribute indexing river segment number. A unified structure of "basin dilution model hierarchical data" is constructed by using time-step linear superposition, empirical weighted averaging, or Bayesian confidence fusion methods. This data takes the time-series dilution concentration change curve of each river segment as the core and includes a comprehensive response of three dimensions: molecular diffusion, turbulent diffusion, and shear dispersion. This supports continuous evolution modeling of watershed pollution load and dynamic load transfer calculation.

[0117] Preferably, step S4 includes the following steps:

[0118] Step S41: Perform mechanism modeling on the region with sufficient attenuation data to obtain pollutant load mechanism result data;

[0119] Step S42: Select areas with a forest area ratio greater than 70% for multiple regression model construction in areas with scarce attenuation data, and estimate forest load to obtain pollutant load statistical results;

[0120] Step S43: Perform Bayesian fusion of pollutant load mechanism results data and pollutant load statistical results data to obtain a mixed load output matrix.

[0121] In this embodiment of the invention, a region with sufficient decay data is first identified, namely, a multi-time period node region with continuous measured concentration, flow rate and historical load records. Based on this, a first-order decay model based on reaction kinetics is introduced. Alternatively, a river-level pollutant migration-transformation equation set (such as the Advancement-Dispersion-Reaction model) can be used, combined with hydrological driving data and diffusion parameters (such as hydraulic radius, flow velocity, and residence time) to simulate the physical mechanism of pollutant concentration evolution, outputting "pollutant load mechanism result data". This data is presented as a pollutant load prediction matrix under spatiotemporal distribution, with fields such as river segment / node coding, time index, and mass flux. In step S42, for areas with scarce attenuation data, the proportion of forest land in the land use data is extracted, and areas with forest coverage greater than 70% are selected. A multiple linear regression model or ridge regression model is constructed by combining variables such as soil type, slope, annual precipitation, and runoff depth. Its expression is generally Y=β0+∑β i X i +, where Y represents the estimated pollution load, X i For land use-related independent variables, β iis the regression coefficient, and is the residual term. Coefficients are fitted using calibration data from historically similar watersheds, and estimation inference is performed on the current target area, outputting "pollutant load statistical results data". In step S43, a Bayesian fusion method is used to weight and merge the two types of results. Let the output of the mechanistic model be the prior distribution P(θ), and the result of the statistical model be the observed likelihood P(D|θ). Then, the posterior distribution after fusion is P(θ|D)∝P(D|θ)P(θ). In specific implementation, Markov chain Monte Carlo (MCMC) sampling can be used to solve for the parameters of the posterior distribution and to estimate the confidence interval of the model output, ultimately generating a structurally unified and numerically stable "mixed load output matrix" to characterize the comprehensive output results of multi-source pollution loads in different regions and under different model application conditions, and to provide quantitative input for subsequent pollution contribution rate decomposition.

[0122] Preferably, step S41 includes the following steps:

[0123] Step S411: Analyze nitrification at temperatures above 15℃ based on sufficient attenuation data to obtain ammonia nitrogen conversion.

[0124] Step S412: Based on the sufficient attenuation data, the denitrification effect was analyzed for DO less than 2 mg / L, and first-order kinetics were used to obtain nitrogen removal data;

[0125] Step S413: Compile the ammonia nitrogen conversion and denitrification data according to a monitoring point density greater than 1 / 100km. 2 Mechanism modeling was performed to obtain data on pollutant load mechanism results.

[0126] In this embodiment of the invention, temperature attribute stratification screening is performed on areas with sufficient attenuation data to extract sample units with water temperatures above 15°C. This temperature threshold is the minimum physiological activity temperature range for ammonia-oxidizing bacteria. Based on the empirical relationship between water temperature and nitrification rate, a temperature-corrected first-order reaction rate expression is used. Where k 20 Here, θ is the reaction rate constant at 20℃, θ is the temperature sensitivity coefficient (typically ranging from 1.03 to 1.07), and T is the measured water temperature. The initial ammonia nitrogen concentration is used to calculate the "ammonia nitrogen conversion rate". In step S412, water bodies with DO concentrations below 2 mg / L are selected to identify potential areas for denitrification and a first-order kinetic model is used. Estimate the denitrification reaction rate, where k dn The rate constant for the denitrification reaction is... The concentration of nitrate nitrogen was determined by analyzing historical monitoring data for k. dnRegional calibration or empirical correction is performed, and the C / N ratio is introduced as a process correction factor for accuracy verification, ultimately generating a "denitrification data" matrix. Step S413 spatially integrates the nitrogen speciation data generated in steps S411 and S412, and combines it with monitoring point distribution information overlaid on the GIS platform to select spatial areas with a monitoring point density greater than 1 / 100km2 as effective areas for mechanism modeling. Within this area, a water quality evolution model is constructed using a set of pollutant transformation equations, with ammonia nitrification and nitrate denitrification as coupled processes. A mechanism model including nitrogen mass balance and temperature-DO dual-factor driven models is constructed, generating "pollutant load mechanism result data". This data is indexed by river segment number and includes time-series nitrogen mass flux, transformation ratio, process rate parameters, and decay factor fields, possessing completeness and consistency to support pollution source analysis and spatiotemporal process coupling simulation.

[0127] Preferably, step S5 includes the following steps:

[0128] Step S51: Perform multi-objective calibration on the mixed load output matrix to obtain multi-objective calibration data; render the multi-objective calibration data into a heat map to obtain a pollution contribution rate distribution map;

[0129] Step S52: Verify the accuracy of the pollution contribution rate distribution map to obtain watershed water pollution simulation verification data;

[0130] Step S53: Construct a qualified report for sediment / nutrients in the Dianchi Lake basin using the simulated verification data of water pollution in the basin.

[0131] In this embodiment of the invention, the "mixed load output matrix" is used as the calibration object, and a parameter optimization problem containing multiple objective functions is constructed. The multi-objective calibration process typically includes typical water quality / quantity fitting indices such as the Nash-Sutcliffe efficiency coefficient (NSE), root mean square error (RMSE), and deviation rate (PBIAS), and introduces multi-source measured concentrations, flow velocities, and load observations as reference data for matching. Using strategies such as the Non-Dominated Sorting Genetic Algorithm-II (NSGA-II) or Pareto front optimization, the key influencing parameters in the Bayesian fusion output matrix are iteratively calibrated until all multiple objective functions reach convergence criteria. The resulting "multi-objective calibration data" includes the parameter set of each sub-basin or HRU unit and its matching quality with the actual observation data. In the visualization stage, the pollution source contribution values ​​after calibration are normalized and mapped to color levels according to spatial encoding. A heatmap rendering algorithm is used to intuitively express the spatial differences in pollution load using color gradients, generating a "pollution contribution rate distribution map," where each pixel in the image maps to a pollution source intensity value. In step S52, the system verifies the accuracy of the simulation results reflected in the heat map against the observed data from the actual monitoring points, mainly using error statistical testing methods such as mean square error (MSE), mean absolute error (MAE), and R². 2 The correlation coefficient, based on the spatial interpolation results and the comparison of point-to-area errors with the observation point data, generates "basin water pollution simulation verification data". In step S53, using the national or regional limits for sediment and nutrient concentrations in the Dianchi Lake basin as the qualification threshold standard, the system performs qualitative and quantitative assessments of the simulation verification results. Combining the contribution rate of each sub-basin and its probability interval of exceeding the limit, a "Dianchi Lake Basin Sediment / Nutrient Compliance Report" is generated, which includes elements such as pollutant type, spatial location, concentration level, and compliance status. This report is output in a structured format and has functions such as overall evaluation, key area identification, and limit comparison analysis, providing direct evidence for basin water environment management.

[0132] Of particular importance, step S51 includes:

[0133] Obtain historical measured data on runoff, sediment, nitrogen, and phosphorus;

[0134] The runoff volume was calculated using the Nash efficiency coefficient based on historical measured data of runoff-sediment-nitrogen and phosphorus, and relative error analysis was performed on sediment concentration to obtain the objective function error analysis data.

[0135] The objective function error analysis data and the mixed load output matrix were simultaneously optimized for hydrological and water quality parameters to obtain multi-objective calibration data; the multi-objective calibration data were then rendered into a heat map to obtain a pollution contribution rate distribution map.

[0136] In this embodiment of the invention, a joint hydrological and water quality observation dataset containing runoff, sediment concentration, and nitrogen and phosphorus concentrations from different time periods and monitoring points is obtained. A standard-format historical measured data matrix is ​​established, with data fields including time index, spatial location, measured values, and corresponding environmental attributes. For runoff rate determination, the Nash-Sutcliffe Efficiency (NSE) coefficient is used as the goodness-of-fit evaluation index, and its expression is: Q o bs,i and Q sim,i These are the observed values ​​and the simulated values, respectively. The average of observed values ​​is used to assess the ability of simulated runoff to reconstruct observed data. In the error analysis of sediment concentration, relative error (RE) or mean absolute percentage error (MAPE) is used to calculate the deviation between observed and simulated values. A tolerance-based error assessment standard is established for highly fluctuating sediment load data, forming the "objective function error analysis data." Next, the system performs joint optimization of hydrological and water quality parameters with the generated "mixed load output matrix." Key parameters, such as runoff time, infiltration coefficient, soil adsorption rate, and nitrogen and phosphorus deposition rate, are adjusted using multi-objective optimization methods (such as weighted least squares fitting, genetic algorithms, or MCMC sampling), resulting in "multi-objective calibration data" with observational support constraints. Finally, in the data visualization stage, spatial coding is used to render heatmaps of the pollution contribution values ​​of each cell. A standard color-gradient mapping model (such as RGB or HSV gradient conversion) is employed, and spatial projection is used to express the pollution source contributions of each sub-basin in a two-dimensional map using color intensity, generating a "pollution contribution rate distribution map." This provides an intuitive and quantitative representation of the spatial distribution of pollutant loads. This process starts with measured data and gradually achieves alignment between the simulation output and the actual situation, providing a data foundation for subsequent simulation verification and governance assessment.

[0137] Of particular importance, step S52 includes:

[0138] The pollution contribution rate distribution map was classified according to the Nash coefficient greater than 0.75 to obtain watershed pollution classification data;

[0139] Based on the watershed pollution classification data, pollution risk control data is obtained by classifying those with a contribution rate greater than 30% as red alerts and the rest as green alerts.

[0140] Pollution risk management data are graded and coded to obtain watershed water pollution simulation verification data.

[0141] In this embodiment of the invention, the simulated pollutant concentration values ​​corresponding to each spatial unit in the pollution contribution rate distribution map are compared point by point with the measured concentration values. The Nash-Sutcliffe Efficiency (NSE) coefficient is used as the criterion for judging the model fitting accuracy. Regions with an NSE greater than 0.75 are extracted as high-fit units, resulting in "basin pollution classification data." This classification data is stored using spatial coding and includes sub-basin number, pollution type, simulated and observed concentration sequences, and NSE value fields. Based on this, the system extracts the pollution contribution rate of each cell in the pollution classification data. If its corresponding contribution value is higher than 30%, it is marked as a "red warning," representing a region with significant pollution risk; the remaining regions are marked as "green warnings," constituting "pollution risk control data." To enhance data traceability and structural integrity, the system introduces a hierarchical coding mechanism. It generates coded strings based on pollution source type (e.g., agricultural, domestic, non-point source), warning level (red / green), and sub-basin number. For example, "R-AG-0508" represents a red warning, an agricultural source, and sub-basin unit 0508, ultimately generating "basin water pollution simulation verification data." This dataset employs a nested combination of spatial indexing and pollution levels, facilitating subsequent model backtesting, policy intervention, and the formulation of zoned governance strategies. It also provides standardized verification samples for analyzing the spatiotemporal evolution trends of pollutants. The entire process emphasizes a coupling mechanism between measured data-driven parameter selection, quantitative contribution grading, and structured output, ensuring consistency between simulation accuracy and risk identification.

[0142] This specification provides a watershed water pollution migration and transformation simulation system based on the MT-DHM model: This system executes the aforementioned watershed water pollution migration and transformation simulation method based on the MT-DHM model and includes:

[0143] The watershed data acquisition and standardization module is used to acquire data from multiple sources in the watershed and perform data preprocessing to generate standard datasets for hydrological response units.

[0144] The vertical runoff and migration correction module is used to perform vertical stratification calculations based on the standard dataset of hydrological response units to obtain vertical analysis data of land runoff; it simulates the migration of land pollutants from the vertical analysis data of land runoff and performs real-time correction with an iteration step of 1 hour to obtain the dissolved / adsorbed load matrix.

[0145] The water system migration and dilution hierarchy simulation module is used to perform push-flow migration calculations on the dissolved / adsorbed load matrix and to perform joint simulation of the dispersed dilution hierarchy to obtain watershed dilution model hierarchy data. The watershed dilution model hierarchy data is then automatically transferred to the nodes of a multi-branch tree to obtain the inter-node attenuation load matrix, which includes regions with sufficient attenuation data and regions with scarce attenuation data.

[0146] The mechanism / statistical modeling and Bayesian fusion module is used to perform mechanism modeling in areas with sufficient attenuation data to obtain pollutant load mechanism result data; to perform statistical modeling in areas with scarce attenuation data to obtain pollutant load statistical result data; and to perform Bayesian fusion of pollutant load mechanism result data and pollutant load statistical result data to obtain a mixed load output matrix.

[0147] The contribution rate calibration and qualified report generation module is used to perform multi-objective calibration on the mixed load output matrix to obtain a pollution contribution rate distribution map; to verify the accuracy of the pollution contribution rate distribution map to obtain watershed water pollution simulation verification data; and to construct a qualified report for sediment / nutrients in the Dianchi Lake watershed using the watershed water pollution simulation verification data.

[0148] The beneficial effects of this invention are as follows: First, it acquires multi-source spatial data encompassing diverse and heterogeneous information, including DEM topographic data, land use data, soil parameters, and measured rainfall and runoff data. Second, it uses data cleaning techniques such as unified projection, time alignment, rasterization, and attribute reconstruction to form a standardized dataset of hydrological response units with consistent structure, laying a high-quality data foundation for subsequent analysis. Third, it constructs continuous hydrological profile data from vegetation interception and soil infiltration to groundwater runoff through vertical stratification calculation. Combined with measured parameters, it simulates and corrects pollutant migration in real time, achieving dynamic load calculation with a 1-hour step size and obtaining a dissolved / adsorbed pollution load matrix. Fourth, it models the spatial transport and attenuation process of pollutants as a multi-branch tree structure using a plug-flow migration and multi-path dispersion dilution model, effectively expressing the load transfer and zonal attenuation characteristics between upstream and downstream nodes. The attenuation data modeling is divided into mechanistic and statistical models. Bayesian fusion technology is used to jointly process structured and unstructured regional load data, significantly improving the robustness and data adaptability of the simulation results. In the model output stage, combining historical runoff, sediment, and nitrogen and phosphorus measured data, and employing multi-criteria model evaluation methods such as Nash efficiency coefficient and multi-objective calibration algorithms, the simulation results are spatially distributed and errors are optimized to ensure the high reliability of the pollution contribution rate distribution. This facilitates pollution risk early warning and compliance verification assessment. Ultimately, the model achieves compliance classification and output of sediment and nutrient indicators at the watershed scale, providing direct data support and decision-making basis for environmental management and pollution control.

[0149] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0150] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A watershed water pollution migration and transformation simulation method based on the MT-DHM model, characterized in that, Includes the following steps: Step S1: Collect data from multiple sources in the watershed and preprocess the data to generate a standard dataset of hydrological response units; Step S2: Perform vertical stratification calculations based on the standard dataset of hydrological response units to obtain vertical analysis data of land runoff; Land pollutant migration simulation was performed on vertical analysis data of land runoff, and real-time correction was performed with an iteration step of 1 hour to obtain the dissolved / adsorbed load matrix; Step S3: Perform push-flow migration calculations on the dissolved / adsorbed loading matrix and conduct joint simulations of the dispersion and dilution hierarchy to obtain watershed dilution model hierarchy data; Automatic load transfer between nodes of the watershed dilution model hierarchy data is performed on multi-branch tree nodes to obtain the inter-node attenuation load matrix, which includes regions with sufficient attenuation data and regions with scarce attenuation data. Step S4: Perform mechanism modeling on the region with sufficient attenuation data to obtain pollutant load mechanism result data; Statistical modeling was performed on areas with scarce attenuation data to obtain statistical results of pollutant load. The pollutant load mechanism results data and pollutant load statistical results data are fused using Bayesian method to obtain the mixed load output matrix; Step S5: Perform multi-objective calibration on the mixed load output matrix to obtain the pollution contribution rate distribution map; verify the accuracy of the pollution contribution rate distribution map to obtain the watershed water pollution simulation verification data; construct the Dianchi watershed sediment / nutrient qualification report using the watershed water pollution simulation verification data.

2. The watershed water pollution migration and transformation simulation method based on the MT-DHM model according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Obtain hydrological dataset from hydrological monitoring station; Step S12: Extract DEM data from the hydrological dataset at a resolution of 30m, perform watershed boundary analysis, and generate sub-watershed delineation maps; Step S13: Extract river network data from the hydrological dataset according to the river section slope, and perform topological relationship parsing to obtain a multi-branch tree node encoding table; Step S14: Divide the hydrological dataset into land use / soil type data according to 1km grids, and perform spatial overlay analysis to obtain the HRU cell attribute matrix; Step S15: Perform Thiessen polygon spatial interpolation on the hydrological dataset at the hourly level to generate sub-basin surface rainfall sequences; Step S16: Construct soil adsorption parameters from the hydrological dataset and perform laboratory calibration to obtain a watershed adsorption parameter table; Step S17: Clean the data from the sub-basin division map, multi-branch tree node coding table, HRU unit attribute matrix, sub-basin surface rainfall sequence, and watershed adsorption parameter table, and format and unify them to obtain the standard dataset of hydrological response units.

3. The watershed water pollution migration and transformation simulation method based on the MT-DHM model according to claim 1, characterized in that, Step S2, which involves vertical stratification calculation based on the standard dataset of hydrological response units, includes: Rainfall data was extracted from the standard dataset of hydrological response units to obtain watershed rainfall data; runoff process analysis was performed on the standard dataset of hydrological response units to obtain land watershed runoff data; Vertical stratification calculations were performed on watershed rainfall data and land watershed runoff data to obtain vertical analysis data of land runoff. The vertical stratification calculations included: calculating the interception capacity of the vegetation interception layer based on a leaf area index (LAI) greater than 2.0 for both watershed rainfall data and land watershed runoff data to generate groundfall rainfall data; and performing infiltration analysis on the vegetation interception layer based on a saturation rate of the upper soil layer greater than 40% for both watershed rainfall data and land watershed runoff data to obtain the infiltration water component data of the land watershed. Vertical data integration of groundfall rainfall data and land watershed infiltration data yields vertical analysis data of land runoff.

4. The watershed water pollution migration and transformation simulation method based on the MT-DHM model according to claim 1, characterized in that, Step S2 involves simulating land pollutant migration from the vertical analysis data of land runoff and performing real-time correction with an iteration step size of 1 hour, including: Obtain the soil permeability coefficient; Based on soil permeability coefficient, the leaching rate of dissolved nitrogen or nitrate nitrogen is calculated from the vertical analysis data of land runoff to generate dissolved migration flux. The amount of adsorbed phosphorus was calculated by analyzing the vertical analysis data of land runoff based on the amount of adsorbed phosphorus stripped when the slope of the soil shear force is greater than 5°. The dissolved state migration flux and the adsorbed state migration flux are merged into adsorption levels and real-time correction is performed with an iteration step size of 1 hour to obtain the dissolved state / adsorbed state loading matrix.

5. The watershed water pollution migration and transformation simulation method based on the MT-DHM model according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Perform plug flow migration calculation on the dissolved / adsorbed loading matrix, and calculate the flow velocity based on the roughness of 0.05 according to the Manning formula to obtain the inflow plug flow data; Step S32: Perform a joint simulation of the dispersed dilution hierarchy based on the inflow propagation flow data to obtain the watershed dilution model hierarchy data; Step S33: Automatically transfer the loads of the multi-branch tree nodes to the hierarchical data of the watershed dilution model to obtain the inter-node attenuation load matrix, which includes regions with sufficient attenuation data and regions with scarce attenuation data.

6. The watershed water pollution migration and transformation simulation method based on the MT-DHM model according to claim 5, characterized in that, Step S32 includes the following steps: Step S321: Perform diffusion coefficient analysis on the inflow propulsion data. ―5 m 2 The molecular diffusion dilution data were obtained by calculating the molecular diffusion dilution data using Fick's law for / s. Step S322: Correct the inflow propulsion flux data for eddy viscosity coefficient with a Reynolds number greater than 4000 to obtain turbulent diffusion dilution data; Step S323: Perform shear flow discretization analysis on the inflow propulsion data with a channel curvature greater than 1.2 to obtain dilution data due to dispersion effect; Step S324: Integrate the molecular diffusion dilution data, turbulent diffusion dilution data, and dispersion dilution data to obtain the watershed dilution model hierarchical data.

7. The watershed water pollution migration and transformation simulation method based on the MT-DHM model according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Perform mechanism modeling on the region with sufficient attenuation data to obtain pollutant load mechanism result data; Step S42: Select areas with a forest area ratio greater than 70% for multiple regression model construction in areas with scarce attenuation data, and estimate forest load to obtain pollutant load statistical results; Step S43: Perform Bayesian fusion of pollutant load mechanism results data and pollutant load statistical results data to obtain a mixed load output matrix.

8. The watershed water pollution migration and transformation simulation method based on the MT-DHM model according to claim 7, characterized in that, Step S41 includes the following steps: Step S411: Analyze nitrification at temperatures above 15℃ based on sufficient attenuation data to obtain ammonia nitrogen conversion. Step S412: Based on the sufficient attenuation data, the denitrification effect was analyzed for DO less than 2 mg / L, and first-order kinetics were used to obtain nitrogen removal data; Step S413: Compile the ammonia nitrogen conversion and denitrification data according to a monitoring point density greater than 1 / 100km. 2 Mechanism modeling was performed to obtain data on pollutant load mechanism results.

9. The watershed water pollution migration and transformation simulation method based on the MT-DHM model according to claim 1, characterized in that, Step S5 includes the following steps: Step S51: Perform multi-objective calibration on the mixed load output matrix to obtain multi-objective calibration data; render the multi-objective calibration data into a heat map to obtain a pollution contribution rate distribution map; Step S52: Verify the accuracy of the pollution contribution rate distribution map to obtain watershed water pollution simulation verification data; Step S53: Construct a qualified report for sediment / nutrients in the Dianchi Lake basin using the simulated verification data of water pollution in the basin.

10. A watershed water pollution migration and transformation simulation system based on the MT-DHM model, characterized in that: The watershed water pollution migration and transformation simulation method based on the MT-DHM model as described in claim 1, wherein the watershed water pollution migration and transformation simulation system based on the MT-DHM model comprises: The watershed data acquisition and standardization module is used to acquire data from multiple sources in the watershed and perform data preprocessing to generate standard datasets for hydrological response units. The vertical runoff and migration correction module is used to perform vertical stratification calculations based on the standard dataset of hydrological response units to obtain vertical analysis data of land runoff; it simulates the migration of land pollutants from the vertical analysis data of land runoff and performs real-time correction with an iteration step of 1 hour to obtain the dissolved / adsorbed load matrix. The water system migration and dilution hierarchy simulation module is used to perform push-flow migration calculations on the dissolved / adsorbed load matrix and to perform joint simulation of the dispersed dilution hierarchy to obtain watershed dilution model hierarchy data. The watershed dilution model hierarchy data is then automatically transferred to the nodes of a multi-branch tree to obtain the inter-node attenuation load matrix, which includes regions with sufficient attenuation data and regions with scarce attenuation data. The mechanism / statistical modeling and Bayesian fusion module is used to perform mechanism modeling in areas with sufficient attenuation data to obtain pollutant load mechanism result data; to perform statistical modeling in areas with scarce attenuation data to obtain pollutant load statistical result data; and to perform Bayesian fusion of pollutant load mechanism result data and pollutant load statistical result data to obtain a mixed load output matrix. The contribution rate calibration and qualified report generation module is used to perform multi-objective calibration on the mixed load output matrix to obtain a pollution contribution rate distribution map; to verify the accuracy of the pollution contribution rate distribution map to obtain watershed water pollution simulation verification data; and to construct a qualified report for sediment / nutrients in the Dianchi Lake watershed using the watershed water pollution simulation verification data.

Images